PNNL Leaders Head to DOE InnovationXLab Summit on Artificial Intelligence
InnovationXLab showcases the assets and capabilities of DOE’s national laboratories
A contingent of PNNL scientists and leaders is heading to Argonne National Laboratory next week to take part in the fourth installment of DOE’s Innovation XLab series – this one focused on artificial intelligence (AI).
InnovationXLab showcases the assets and capabilities of DOE’s national laboratories, bringing together the labs’ innovators and experts with their counterparts in industry, universities, and other institutions potentially interested in the technologies created and developed at the labs.
PNNL scientists will feature six of their technologies, a sampling of the lab’s dozens of projects in the realm of AI, at the event Wednesday-Thursday, Oct. 2-3. In addition, PNNL Director Steven Ashby will moderate a panel discussion focused on the role of AI for energy grid optimization.
DOE Director Rick Perry will deliver the keynote address.
The PNNL technologies featured at XLab touch upon many areas of our daily lives, including energy production, transportation, and safety and security. Here are the PNNL technologies, most of which are available for licensing.
TRAST is a tool that helps electric utilities plan for emergencies and other events. The software, named the Transformative Remedial Action Scheme Tool, includes machine learning algorithms that create, analyze, and validate scenarios to help utilities plan for and prevent emergencies. The tool enhances ways to identify and evaluate complex scenarios; its accuracy and power allow engineers to operate the grid more reliably, more efficiently and at lower cost than they otherwise could.
VitalTag keeps allows first responders to keep close tabs on patients and others involved in a mass casualty incident, offering life-saving efficiency and clarity in times of turmoil. The technology includes a disposable sensor suite that measures a person’s EKG, body temperature, respiratory rate, blood oxygen level, and other vital signs. That information is sent wirelessly to responders, allowing them to closely monitor the status of many people at one time – for triage purposes, for example, or as an immediate alert to a sudden change in conditions.
The Acoustic Gunshot Detector detects a gunshot indoors instantly, providing information on the location and the type of weapon. The information can help first responders take swift action, initiating emergency and law enforcement measures. Using machine-learning capabilities, researchers trained the sensors to not only identify a gunshot but also the caliber of the weapon used. The technology, which can identify a gunshot with 99.99 percent accuracy, is now deployed in schools and other settings.
Identify n-D is technology that can locate just about anything on the Earth’s surface in a fraction of a second – a specific land parcel, property boundaries for homeowners, a cave, the restrooms in a building on another continent, or the HVAC system in a corporation’s holdings across the globe. The system creates a string of code that is unique to every location on the planet. Identify n-D can link seamlessly with other systems, such as industrial tracking systems, building permit software, flood maps, and weather data, to yield insights such as the specific dangers posed by a looming hurricane or to explain the location of misplaced equipment.
PNNL’s work characterizing the quality of ultrasonic bonds relies on machine learning and sophisticated acoustic sensors that monitor weld quality, an important feature for any piece of equipment or device that has been built using more than one piece of material. The system sends acoustic signals during the welding process to monitor the quality of the weld in real time – during fabrication, when the information is most valuable. The technology increases the efficiency of the manufacturing process and works with a variety of metals, including aluminum, copper, platinum, and titanium.
Sharkzor is an AI-based system that helps users classify images much more quickly than other machine-learning methods. It differs from similar systems in a fundamental way: While most machine learning systems rely on human input only when the algorithm is stumped, Sharkzor keeps humans in the loop all along and calls upon human interaction at the outset of a challenge. This reduces the amount of information needed to sort complex images and speeds the process dramatically. The powerful AI program, with a touch of humanity, results in a system that can classify complex sets of images much faster than programs currently in use.
Published: September 27, 2019